1 comment

[ 3.1 ms ] story [ 14.7 ms ] thread
I suspect the preference is functional rather than aesthetic.

A web UI is sandboxed, whereas the terminal has access to local context. If I am debugging a Celery worker or a Postgres migration, I need the agent to be able to execute the code it generates and read the error logs directly from the container.

You lose all that composability in a browser. The terminal is just the most efficient way to bridge the gap between generation and execution.